Stereotypical Bias Analysis On Crows Pairs
평가 지표
Age
Disability
Gender
Nationality
Overall
Physical Appearance
Race/Color
Religion
Sexual Orientation
Socioeconomic status
평가 결과
이 벤치마크에서 각 모델의 성능 결과
모델 이름 | Age | Disability | Gender | Nationality | Overall | Physical Appearance | Race/Color | Religion | Sexual Orientation | Socioeconomic status | Paper Title | Repository |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| GPT-3 | 64.4 | 76.7 | 62.6 | 61.6 | 67.2 | 74.6 | 64.7 | 62.6 | 76.2 | 73.8 | OPT: Open Pre-trained Transformer Language Models | |
| GAL 120B | 69 | 66.7 | 51.9 | 51.6 | 60.5 | 58.7 | 59.9 | 51.9 | 77.4 | 65.7 | Galactica: A Large Language Model for Science | |
| LLaMA 65B | 70.1 | 66.7 | 70.6 | 64.2 | 66.6 | 77.8 | 57.0 | 70.6 | 81.0 | 71.5 | LLaMA: Open and Efficient Foundation Language Models | |
| OPT-175B | 67.8 | 76.7 | 65.7 | 62.9 | 69.5 | 76.2 | 68.6 | 65.7 | 78.6 | 76.2 | OPT: Open Pre-trained Transformer Language Models |
0 of 4 row(s) selected.